Dual Neural Network for Energy-Efficient Signal Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Artificial neural networks used in microcontrollers for detecting events in physical signals consume significant energy, which is a concern for devices with limited battery life, and simplifying these networks to reduce energy consumption often compromises detection performance.
Innovation Solution
Implementing a dual neural network system where a reference neural network is used for high-probability event detection and a simplified auxiliary neural network is used for low-probability detection, with the auxiliary network being faster and less energy-intensive, allowing for reduced energy consumption while maintaining monitoring capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a reference artificial neural network is used for detecting events in physical signals, then detection performance is improved, but energy consumption increases
Solution Approach 1:
The patent applies dynamics by making the neural network configuration adaptive based on detected event probability. When an event is detected with high probability, the reference neural network is activated to ensure accurate detection. When event probability is low, the system switches to a simplified auxiliary network or stops detection to conserve energy. This dynamic switching resolves the contradiction between maintaining high detection performance and reducing energy consumption.
Solution Approach 2:
The patent applies local quality by using different levels of neural network complexity in different operational contexts. The reference neural network with full complexity is deployed only when needed (high probability events), while a simplified auxiliary network with reduced complexity is used for low-probability events. This localized application of computational resources optimizes both detection performance and energy consumption.
2Use of energy by moving object
If the artificial neural network is simplified to reduce calculations, then energy consumption is reduced, but detection performance deteriorates
Solution Approach 1:
The patent applies partial action by implementing a simplified auxiliary neural network that performs only the essential detection function with reduced computational complexity. This partial implementation is sufficient for low-probability events where high accuracy is not critical, thereby reducing energy consumption while maintaining acceptable detection performance for those specific conditions.
Solution Approach 2:
The patent segments the detection system into two distinct neural networks with different complexity levels. The reference neural network handles high-probability events requiring full accuracy, while the auxiliary neural network handles low-probability events with reduced computational requirements. This segmentation allows each network to be optimized for its specific operational context, resolving the contradiction between simplification and performance.
3Use of energy by moving object
If the artificial neural network stops execution during certain periods, then energy consumption is reduced, but detection capability is lost during these periods
Solution Approach 1:
The patent implements periodic action by continuously monitoring event probability and switching between different neural network modes based on detected conditions. Rather than stopping detection entirely, the system periodically adjusts its computational intensity - using the full reference network when events are detected and switching to simplified mode during low-probability periods, maintaining detection capability while reducing average energy consumption.
Data Source
AI summary
According to one aspect, a method is provided for detecting events or elements in physical signals, including at least one implementation of a reference artificial neural network, at least one implementation of an auxiliary artificial neural network distinct from the reference artificial neural network. The auxiliary artificial neural network being simplified relative to the reference artificial neural network. At least one assessment of a probability of presence of the event or the element by the implementation of the reference artificial neural network or by the implementation of the auxiliary artificial neural network, where the reference artificial neural network is implemented when the probability of presence of the event or the element is greater than a threshold, and wherein the auxiliary artificial neural network is implemented when the probability of presence of the event or the element is below the threshold.